Seeing as to how the original plan was 2 separate boards, perhaps another option is to do so with slightly different partitioning. Obviously this adds to assembly complexity but with board pricing based strictly on real-estate without set-up costs...
I'm thinking putting the MCP1661 circuit on a tiny 2-layer board allowing the backside ground plane. It would essentially be like that 5V switching regulator packaged into a TO-220 you use on the Super Chuffer. So a 3-pin interconnect to the main board. This sub-circuit would then be another "tall" thru-hole component that might free up enough real-estate to use the small-board Keystone method. That is, I assume you have not closed the doors to the peanut gallery... 
